摘要
为了使月球探测器自主的在月球表面进行软着陆,应对着陆区域的地形进行三维重构的方法,获得着陆区域的地形描述。提出了一种通过运动获得长基线的立体视觉的三维重构方法,针对长基线立体视觉方法所存在的问题,通过设计一系列的算法,完成了图像特征点选取和匹配,估计相机在不同位置的相对旋转和位移;对立体图像对进行校正,获得稠密的视差图;通过视差图进行三维重构,生成着陆区域的DEM。根据月球地形的特点建立了软件仿真平台,并且在仿真平台的基础上对文中的算法进行验证。仿真的结果表明该方法可以有效的应用于月面地形的三维重构。
In order to make lunar lander autonomous safely landing on the moon, the terrain of landing region should be rebuilt, and the description of landing site terrain should be generated. A long base-line stereo vision 3D reconstruction method through motion is proposed. Aimed at the problems of long base-line stereo vision, a series of algorithms is designed to select and match feature points; estimate the relative rotation and translation between different positions; rectify stereo image pair; get the dense disparity image; perform 3D reconstruction method through disparity image, and generate the DE M( Digital Elevation Map) of landing region. According to the characteristic of lunar terrain, a software simulation test-bed is built, and on witch the algorithms are tested. The results of simulation show that this method can be used in 3D reconstruction of lunar terrain effectively.
出处
《宇航学报》
EI
CAS
CSCD
北大核心
2007年第4期966-971,共6页
Journal of Astronautics
基金
国家自然科学基金重点资助项目(70475027)
关键词
立体视觉
长基线
图像校正
三维重构
Stereo vision
Long base-line
Image rectification
3D reconstruction